1 Introduction

The morpho-syntactic processing of the addressee received increased attention in recent studies that focus on allocutive or on honorific elements merged in the clause. In this respect, dedicated particles or verb endings mark not only the biological traits of the addressee (i.e., gender or age with allocutives; see Miyagawa 2012; Haddican 2018; Zu 2018; Alok 2021 a.o.) but also the relative position of the interlocutors in a social hierarchy, or the closeness of their relation (see the honorifics in Portner et al. 2019 a.o.). Crucially, the type of social relation encoded in the clausal domain varies from one language to another, hence the need to investigate the underlying mechanism that allows for the syntactic mapping of this information while also providing room for morphological and lexical variation in its implementation.

While the encoding of social relations through allocutive agreement in the clausal domain attracted the attention of a number of scholars working on a variety of languages (see the contents of this volume), less attention has been paid to the encoding of similar information within the nominal domain, particularly in the forms of direct addresses that qualify as Vocative Phrases (VocP), in which the speaker overtly identifies the addressee and their social relation (Zwicky 1974; Stavrou 2014 a.o.). In this respect, previous studies (Arsenijević 2007; Abdulaze & Ludden 2013; D’Avis & Meibauer 2013 a.o.) notice that, cross-linguistically, dedicated particles (some of which arise from the grammaticalization of pronouns) merge in VocPs to express the speaker’s disapprobation for an addressee with a lower social status.

Following this line of research, this paper focuses on two questions: (i) What is the mechanism that encodes disapprobation in VocPs, and (ii) how does that extend to the complex VocP structures known as reversed vocatives?

Most data come from standard modern Romanian, where VocP displays morphological complexity on an apparently optional basis: the vocative noun may or may not display vocative case inflection (voc as in (1a)), definite articles (‘the’ as in (1b)), and/or dedicated vocative particles (voc-p as in (1c)), or be substituted by a vocative particle (1d).

In addition, the kinship rank of the speaker may also be lexicalized through reversed vocatives, as in (2).

In (2), mamă ‘mother’ indicates the speaker’s kinship status, and in the name of this kinship the mother feels entitled to express maternal baby endearment toward an adult son. Hence, the vocative belittles the social standing of the addressee, notwithstanding the positive feelings (adult sons feel embarrassed if that is witnessed by strangers).

The working hypothesis is that VocP contains a formal inter-personal [i-p] feature whose checking and valuation varies at intra- and cross-linguistic levels, as shown in Hill (2014). The variations are conditioned by the options for feature valuation, which may or may not take place; if it does, the inquiry should clarify to what extent it captures the social relation between interlocutors. Along these lines, the argument is that unvalued [i-p] does not provide information about the relation between speaker and addressee, beyond what may be inferred through the principle of compositionality from the sentence. On the other hand, the valued [i-p], by means of dedicated particles, vocative case marking or grammaticalized lexical items, explicitly encodes, at various degrees, the social status of the addressee, as well as the personal relation between the interlocutors.

Building on this proposal, this paper argues that reversed vocatives as in (2) arise from an instance where the [i-p] value is fixed for the identification of the kin status. Unlike the allocutive markers that merge in the clausal domain, which generally signal biological gender and/or honorificity and deference, reversed vocatives serve to systematically assert the speaker’s kinship rank superiority over the addressee. Honorific particles such as seen in the clausal domain in Japanese (Oguro 2015; Miyagawa 2022 a.o.) or Korean (Portner et al. 2019 a.o.) do not occur within VocP despite the presence of [i-p], for which such particles or their equivalent should qualify as goals. As a general empirical observation, particles that occur in vocative phrases do not honor the addressees but rather demote them. This observation does not entail that anti-honorifics may never be found as allocutive markers in the clausal domain of some language (in fact they are, as mentioned for Telugu in Slocum 2016), or that deference cannot be expressed through vocatives (e.g., your majesty!). The point is that, although there is no theoretical constraint for the distribution of dedicated particles, the cross-linguistic tendency is to use the vocative phrase rather than the clausal domain for encoding social demotion or disapprobation.

The paper is organized as follows: Section 2 summarizes the formalisation of vocatives as structures that map the addressee. Section 3 summarizes the formalisation of the information on the social status and presents the basic structure of VocP in (14), such as proposed in Hill (2014). Section 4 builds on (14) to analyse the reversed vocatives as in (21). Section 5 provides a cross-linguistic overview of the use of vocatives as the medium for encoding disapprobation. Section 6 concludes the paper.

2 Formal features in vocative phrases: [2p]

This section summarizes the formal analyses of vocative phrases such as available in current literature. Briefly, building on the observations in Moro (2003), Hill (2007); Espinal (2013) and Stavrou (2014) conclude that vocatives involve an extra-layer of structure above the DP/NP, which accounts for the direct address interpretation of these structures, as opposed to nominal phrases that belong to reports or narratives. Theoretical justification for this extra-layer comes from previous studies, as follows:

Fink (1972) pointed out that any vocative noun receives a 2^nd person reading, since it identifies the addressee. This is in contrast with non-vocative nouns, with intrinsic 3^rd person features. Bernstein (2008a; b) formalizes this observation by arguing that the definite article, intrinsically endowed with a third person feature, checks and values the underspecified [person] feature of D as 3^rd person; hence, vocatives are possible only if the definite article is eliminated. Instead, [person] on D is checked by N-to-D, as in Longobardi (1994), and it is pragmatically or deictically valued for 2^nd person (henceforth [2p]). This contrast between vocative and non-vocative nouns is captured by binding restrictions, as in (3), where the fronted nouns allow for different co-reference with the verb arguments according to whether they are vocatives or topics.

Hill (2007) adopts the line in Longobardi (1994) but points out that N-to-D is not sufficient to account for Romanian vocatives that not only maintain the definite article but also add a vocative case ending to it (e.g., in (3c) above). So an extra-layer over DP is proposed, as RoleP, which is relabeled VocP in Haegeman & Hill (2010); Espinal (2013); Hill (2014); Stavrou (2014). Furthermore, following D’Hulst et al. (2007), Hill (2007; 2014) shows that not only NPs may qualify as vocatives, through head-to-head movement (as implied in Longobardi 1994 and argued for in Espinal 2013), but also DPs (e.g., [My wonderful daughter], how are you?’). Formally, the proposal is that, in vocative structures, the person feature comes as uninterpretable [2p], and it is associated with a different head, higher than D, which allows this feature to be checked either by DP or by N. [2p] probes D or N, whichever comes first, and either head-to-head or phrasal movement to Spec takes place, yielding the hierarchy VocP > (DP) > nP.

Another peculiarity of vocative phrases to be accounted for was that many languages display dedicated vocative particles that have an intrinsic 2^nd person reading, and which may either alternate or cooccur with a vocative N/NP, as shown in (1c) and further in (4) (a cross-linguistic list of such particles is available in Hill 2014).

These particles are obligatory for a noun to be interpreted as a vocative in some languages (i.e., Bantu affixes; yaa series in Arabic), but not in others. That is, [2p] can be checked only by such a particle in some languages, but by N/DP raising in other languages. This raises the question of why vocative particles are also seen in languages where [2p] is checked by N/DP, e.g. Romanian in (4) above. The answer in Hill (2014) is that [2p] is bundled with another feature, identified as [i-p], that allows for the mapping of the social standing of the interlocutors.

3 Formal features in vocative phrases: [i-p]

In (4a), VocP verifiably involves N-to-Voc triggered by the uninterpretable [2p] feature of Voc, since the noun displays the vocative case ending. The morpheme bre is a dedicated vocative particle (i.e., it does not occur in any other environment) inherently endowed with an addressee reading (i.e., it has an interpretable [2p] feature). However, the particle agrees without competing with the vocative marked noun that checks the [2p] of Voc. Notably, bre is optional, as indicated by the brackets around it, while N-to-Voc is obligatory. The option for this particular particle arises when the speaker intends to signal a familiar and slightly condescending address, restricted to the elderly, and which would be inappropriate for an addressee of his/her age bracket. Hence, VocP has not only [2p] as a formal feature, but also a feature that syntactically encodes (i.e., triggers the merging of particles for feature checking operations) the inter-personal [i-p] type of relation between speaker and addressee.

There are tests that confirm the morpho-syntactic status of the vocative particle, as opposed to considering it a post-syntactic phenomenon (Hill 2007; 2014). Incidentally, the tests also demonstrate that the particle merges within VocP, as opposed to merging on the clausal spine, concurrently with VocP. An overview of these tests follows in the remainder of this section.

4 Reversed vocatives

Reversed vocatives (noticed in Robertson Rieschild 1998) are direct addresses that encode not only the identity of the addressee, but also the kinship relation of the speaker and the addressee, as in (15): the speaker spells out the kinship rank that gives him/her the authority to advise or command the addressee, or to emotionally dominate the situation (the speaker steals the spotlight) or blackmail the addressee (through endearments) into doing something. Thus, these constructions are further examples of vocatives that assert the speaker’s privileged standing in the conversation. The construction occurs in a wide range of languages, but is especially productive in Arabic dialects and in some Balkan languages.

In (15), the vocative seems to occur in doubles: while the addressee is identified by name (Dane in (15a)), by the vocative particle (ja in (15b)) or by the inflection of the imperative verb in (15c), the speaker is also signaled in terms of a kinship relation to the addressee, i.e. ‘mother’ or ‘dad’. The reading does not allow us to classify ‘mother’ or ‘dad’ as exclamations (i.e., non-address), since they fully indicate the participant (i.e., first person) versus the generic use of ‘mother’ (i.e., third person) in an exclamation, e.g., for the general purpose of complaining (e.g., Mamă ce de lume! ‘mother what a crowd’/’Mamma mia, so many people!’). Thus, by interpretation and deictic value, this noun qualifies as a vocative. The kinship noun reverses the attention from the identification of the addressee to the speaker’s kin rank, which becomes the spotlight. Due to the kin rank, the speaker can issue orders, pleas, endearing requests and so on. Notably, this social relation is personal and individualistic: neither the addressee nor the kin rank may come in plural or in coordination phrases.

Crucially, the speaker is never identified with regard to his/her name but only with regard to the kinship standing. Thus, the first observation is that the relevant nouns encode a social relation, not the [speaker] feature (i.e., not [1p]) that can be seen in the speech act area of the clausal structure. That is, this vocative construction cannot substitute a name to the kinship noun, nor can it refer to any possible speakers (e.g., a teacher, a stranger, a friend etc.), but only to those who qualify for a kinship relation. This is a cross-linguistic generalization, variation arising only in the composition of the lexical set: while ‘mother’ and ‘father’ occur across the board, other kinship nouns may vary from one language to another (e.g., ‘aunt’ does not occur in Romanian but it does in Arabic dialects, Georgian, Armenian, Albanian; Braun 1985: 83; Boeder 1989: 14).

When it comes to intonation, reversed vocatives form one prosodic unit, with one high pitch. When both the addressee and the kinship nouns are present (15a, b), the pitch falls on the addressee noun/particle; when the kinship noun occurs by itself (15c), it carries the pitch.

Reversed vocatives display morpho-syntactic restrictions with respect to the form of the lexical items and to the permissible word order. First, the word order addressee > kinship noun is fixed within the prosodic unit, as in (16a), or else ungrammaticality ensues, as in (16b).

Second, as long as the two items belong to the same prosodic unit, adjacency is obligatory. With two prosodic units, the vocative can be split, as in (17a), on the condition that the hierarchy is maintained; that is, the addressee c-commands the kinship noun or else ungrammaticality ensues, as in (17b).

In fact, (17b) triggers a misleading analysis: ‘mother’ is parsed as the subject of the imperative, instead of the intended ‘Dan’. This indicates that the kinship noun must be kept away from 2^nd person features, and its licensing somehow depends on the addressee noun.

Third, the complexity of the lexical items involved is very reduced. The kinship noun cannot be modified, as shown in (18a). The licit version comes as a frozen bare form. An exception occurs for the pragmaticized use of ‘dear’, as in (18b). However, ‘dear’ is an attribute of the addressee, not of the speaker (i.e., ‘dear son’ not ‘dear mother’), despite its insertion in the kinship phrase. Inflectional endings are disallowed on the kinship noun, as in (18c), so the noun is bare, unless it comes as a diminutive, in which case the definite article and the vocative case marking are obligatory, as in (18d). The inflected forms of the kin noun are generally preferred when the addressee is non-lexical (18d).

The examples in (18) indicate that the kinship noun has a DP structure, since an adjective (18b) and a definite article (18d) can be merged with N (although the adjective modifies the speaker).³ The fact that the inflected form is dis-preferred indicates a path of grammaticalization, whereby the XP is gradually reanalyzed as an X item (van Gelderen 2011).⁴ The target stage is not yet attained at this time, so the kin noun may have a mixed DP/D/N status. In sum, the kinship noun tends to be bare. However, its syntactic status is not that of a head, since variations are possible in which N has an adjective like ‘dear’ in its specifier, which modifies the addressee noun when the latter moves through Voc2 as shown below in (21). The same kinship noun may also display definite articles, as in (18d), which indicates that it can also project to DP. Hence, the bare use is only the default, not the exclusive occurrence of the kinship noun in reverse vocatives, and the possibility of having the other options indicates that the kinship noun is rather phrasal (vs head).

On the other hand, the addressee noun is systematically bare with reversed vocatives. DP/NPs rule out the construction, as shown in (19b) (hence the placement of ‘dear’ in (18) with the kinship noun, which is phrasal, instead of being merged with the vocative noun).

In (19a), a regular VocP contains a DP that identifies the addressee (DP moved to Spec, VocP). The DP structure is signalled by the presence of the possessive pronoun, which merges only in DP fields. The same DP rules out the reversed vocative in (19b). However, if the possessive pronoun is deleted, as in (19c), the reversed vocative is grammatical, with or without the vocative case ending on the addressee noun. This indicates that the addressee noun systematically undergoes head (versus phrase) movement when reversed vocatives apply.

As a final observation, for a vocative phrase to qualify as a reversed vocative, the kinship noun is obligatorily present. On the other hand, the addressee noun is not obligatorily lexical (e.g., (18d)), yet the expression is unmistakably interpreted as a direct address to a person other than the one indicated by the kinship noun. Along these lines, reversed vocatives are built on the basis of the same feature set as in regular vocatives ([2p] and [i-p]), but with a markedness component that has to be identified.

The inquiry into the markedness that generates a reversed vocative must establish, first, the domain in which the triggering formal feature occurs. So far, the observations hint to the nominal domain, in the vicinity of vocative nouns. However, one may object that the triggering formal feature could rather be associated with the clausal domain (e.g., bundled with [speaker] or [addressee]) and require the same adjacency effects in relation to VocP merged to check the [addressee] feature. That would be a reasonable assumption, since honorific particles, also reminiscent of social ranking, were shown to occur as sentence final particles (Portner et al. 2019). The following observations eliminate that hypothesis:

• – If the kinship DP merged separately from VocP on the clausal spine, it should not be able to optionally carry the vocative case ending, or be modified by adjectives agreeing with the addressee (i.e., VocP is an impenetrable phase for inside agreement), contrary to the fact.

• – The kinship noun provides information about the speaker, so it is expected to merge in a speaker phrase if a clause hierarchy is assumed. First, that would yield the wrong hierarchy, since the speaker is mapped higher than the addressee on the clausal spine (Speas & Tenny 2003; Portner et al. 2019; Miyagawa 2022 a.o.), which would rule out reversed vocatives, where the kinship noun is lower than the addressee noun. Second, a cross-linguistic generalization in formal studies is that speakers are silent even when the mapping of speech acts is otherwise visible (Speas & Tenny 2003; Sigurðsson 2004; 2011; Miyagawa 2022 a.o.). The merging of the kinship noun as the spellout for the [speaker] role would be an exception to this rule, and an ad-hoc solution with no independent support.

• – The type of movement to VocP (i.e., phrasal or head-to-head) should not be constrained by the presence of a kinship noun (contrary to the fact), if the kinship noun is external to the nominal phase.

• – Independent merging on the clausal spine predicts that speech act particles should be able to separate the two items, which is not the case, as shown in (20c), contrasting with the optional placement of the speech act particle in (20a, b), where either relevant item occurs separately. This is a constituency test, leaving no doubt that the two items involved in the derivation of reversed vocatives belong to the same constituent.

Once the clausal spine is excluded from consideration, the VocP domain remains the alternative for the generation of reversed vocatives. Building on the analysis of VocP in (14), that is, two formal features that map the addressee semantics as [2p] and the social relation as [i-p] bundled on Voc (see14)), it follows that reversed vocatives involve a further manipulation of the [i-p] feature that triggers the merging of a kinship noun. In these constructions, the trait that makes the [i-p] feature different from a regular vocative is its strict specification; that is, while in regular vocatives [i-p] is underspecified and can be checked and valued by a variety of items (generating a variety of familiar and/or disapprobation readings), in reversed vocatives [i-p] is strictly reserved for kinship ranking. Hence, [i-p] is reanalyzed as [kin], and VocP is marked.

Along these lines, the markedness of a VocP that supports a reversed vocative arises from the dissociation of [2p] and [kin]: the two features do not bundle on Voc, as in (14), since N-to-Voc, vocative particles that may check [2p] do not qualify to check [kin]. Instead, a noun with kinship semantics is introduced, in a split VocP structure, as in (21).

In (21), [2p] and the [kin] version of [i-p] are mapped to separate Voc heads. The noun spelling the addressee undergoes N-to-Voc, and obligatorily checks Voc1 for [2p], which is the feature that qualifies VocP as an argument in the speech act projection at the left periphery of clauses. That is, due to this feature, VocP becomes visible to the [addressee] role feature mapped on the clausal spine (e.g, as AddresseeP in Miyagawa 2022). Hence, this feature is necessarily located in the highest Voc head: only the item merged at the edge of VocP1 is visible to the [addressee] probe from the clausal spine. This is sufficient to justify the constraint on hierarchical order within the reversed vocative, as signaled for (17b), i.e., the vocative noun is higher than the kinship noun. On the other hand, the kinship noun merges directly in Spec, VocP2 and checks and values [kin]. Incidentally, due to the Spec-head agreement between the vocative noun and the kinship noun in VocP2, the Vocative Case marking may be shared by the two nouns at PF (e.g., Dane mămico ‘Dan.voc mother.voc’ is possible for some speakers).

Notably, (21) conforms to the general property of heads to split when the features associated with them are (for various reasons) checked by different lexical items. In this particular case, Voc splits when [i-p] comes specified as [kin], but remerges when [i-p] is underspecified and can be checked in a bundle with [2p]. Similar split/remerging derivations have been noticed with most functional heads, especially in the process of language change (e.g., in Hill & Alboiu 2016 for Romanian), so (21) is not structurally exceptional.

A note on vocative particles is in order: These particles are not excluded from (21), although [kin] is checked by a different item. The familiarity that the particle introduces in regular vocatives is already present in (21), since the kinship noun has an inherent specification in that respect. In (21), the particle has the effect of enhancing the speaker’s eagerness in the negotiation with the kin addressee. This may indicate a cumulative valuation of [kin], or alternatively, a separate bundling of [kin] with an underspecified [i-p]. This type of cooccurrence is infrequent, but grammatical, and may signal a further markedness development within VocP.

Technically, the DP status of the kinship item (see (18d)) and its merge in a Spec position blocks the DP movement of the addressee noun to Spec, VocP1, but it does not interfere with its head-to-head movement to Voc1; hence, the restriction on vocative noun modification illustrated in (19b). Since this configuration leaves Spec, VocP1 available, a vocative particle may directly merge at that location; hence, the optional presence of the particle in (16)/(21). The local Spec-head agreement that obtains when the addressee noun moves through Voc2 explains the possible inflectional copy of the vocative case ending. Although both the addressee and the kin nouns may display this ending (18d), the default option is to place it only on one of them, preferably on the addressee noun.

Cross-linguistic variation with reversed vocatives is superficial: it concerns the set of kinship nouns and inflectional options (presence or absence of vocative case endings, of vocative particles, definite articles or possessives), but there is no evidence that would challenge the underlying structure in (21).

5 Cross-linguistic perspective

This paper provided mostly Romanian data to argue that the default domain for encoding disapprobation or the speaker’s superior ranking over the addressee is the vocative phrase rather than the clausal field dedicated to the mapping of the conversational pragmatics. However, as shown in this section, observations along these lines abound in the literature on vocatives, although the point hasn’t been nailed down so far regarding the contrast with the use of the clausal space for the default mapping of honorifics. For more clarity, this paper does not claim that the expression of respect for higher ranks is not possible through vocative phrases (actually, the opposite is stated), but that speakers who intend to express disapprobation or social superiority of some kind would more likely than not opt for a VocP structure to do that.

In this respect, the grammaticalization of the pronoun ‘you’ to a vocative particle is well attested as a means of encoding disapprobation especially in languages that do not display dedicated vocative particle or articles to implement it, as we saw for Romanian in section 3. English you idiot! (versus genuine #you hero!) is discussed in Arsenijević (2007), and similar grammaticalization, with similar interpretive effects, are pointed out for German du and Swedish din (D’Avis & Meibauer 2013). The Georgian grammaticalization of ‘you’ ends up with morphologically shorter forms (i.e., še < šen (sg); tkve < tkven (pl)) and serves a disapprobation function as well: “It has to be mentioned that the adjectives and nouns in these expressions almost always have a negative (pejorative) connotation” (Abuladze & Ludden 2013: 28). In Romanian, the grammaticalized tu ‘you’ serves for expressing equality or camaraderie, while articles and/or case marking are reserved for disapprobation. Formally, the grammaticalized ‘you’ becomes a vocative particle, on a par with Rom. bre illustrated in (1), and merges in VocP as indicated in (14) for other vocative particles.

Reversed vocatives also occur in various languages, especially in the Middle East and the Balkans. For example, Georgian, illustrated in (22), shows the same use of kinship nouns as seen in Romanian.

Boeder (1989) explains that Georgian reversed (inversed for her) vocatives are forms of address used only between members of different generations, but not within one generation. The use of da ‘sister’, genacvalos or dzma ‘brother’ are possible in Georgian (not in Romanian, though) as long as these siblings are older than the addressee. Again, the upper hand of the speaker over the addressee is crucial in these constructions, as the addressee must be made aware of his/her inferior ranking, irrespective of how endearing the address may be.

Formally, nothing blocks the mapping of rank inferiority (anti-honorifics) or disapprobation in the clausal domain, and vice-versa for the nominal domain. However, the preference for vocatives as the medium for expressing disapprobation or other anti-honorific sentiments is a cross-linguistic fact.

6 Conclusions

This paper looked at the ways vocative phrases encode the expression of inter-personal relations between speaker and addressee. The basic assumption was that the derivation of vocative phrases relies cross-linguistically on the computation of two features: [2p] and [i-p]. With respect to [2p], which encodes the addressee semantics, variation arises mainly from the option for feature checking mechanisms (i.e., +/– movement of phrases or heads) and sometimes from lexical choices (e.g., nouns or dedicated vocative particles). On the other hand, [i-p], which encodes the inter-personal relation between speaker and addressee, leaves room for more variation insofar as the intrinsic properties may also vary, from underspecified to rigidly specified for a certain type of social relation. This variability accounts for the possibility of reversed vocatives in addition to regular vocatives in the same language.

The question to which this discussion is relevant concerns the identification of the phrasal domain to which information on the social relations of the two discourse participants are mapped as formal features. Formality and honorifics were shown to be expressed in the clausal domain in various languages, either through the merging of dedicated morphemes or through the manipulation of verb inflection (e.g., as sentence final addressee particles in Portner et al. 2019). Romanian makes no exception, as it displays the plural form for the singular reading of the 2^nd person in expressions of formality and respect.

By contrast, this paper pointed out that the type of social relations expressed in the nominal domain by means of dedicated morphemes or case inflection within vocative phrases systematically range from informality to total disapprobation. This was shown to take place in Romanian vocative phrases by the intervention of vocative particles, vocative case marking or definite articles. Therefore, a contrast arises with respect to the mapping domains for honorifics and anti-honorifics by means of functional (versus substantive) lexical items: the former favour the clausal spine, whereas the latter abound in vocative phrases.

Abbreviations

dim = diminutive; f = feminine; imp = imperative; int = interjection; io = indirect object; nom = nominative case; opt = optative; part = speech act particle; refl = reflexive pronoun; pl = plural; sbjv = subjunctive; sg = singular; voc = vocative case; voc-p = vocative particle.

Competing interests

The author has no competing interests to declare.